5 
ac m 
F Sl (STR 
Il 
) Iga 
  
  
43 
A supervisor has an excellent opportunity to evaluate the comple- 
teness of the digitizing by inspecting the trace. 
Using this technique one could expect normal production to be 
about 1 200 points per hour. 
Specifications for a DTM 
  
A peculiar aura seems to be associated with the electronic compu- 
ter and the answers that one gets from it. There is a tendency to 
accept computer generated products from the DTM as "gospel" with 
little concern as to the validity of that data even to the extent 
that ones common sense is overridden. What has been needed is an 
accuracy specification for the DTM just as we have for topographic 
maps since we need to know the validity of products generated from 
a DTM. 
One of the problems in evaluating a DTM is that there are numerous 
algorithms developed for the computer to interpolate elevations 
from a DTM in a specific way. It follows that one should be able 
to not only evaluate the DTM, but the system of computer algorithms 
that are to be used with a specifically structured DTM. 
I know of several instances in which the only way available to test 
the DTM was to visually compare computer generated contours with 
hand-drawn contours from a stereoplotter. At best, this is a poor 
way to test the DTM. 
More often than not, the digital data for a DIM is gathered using 
a digitized stereoplotter of some kind. Here is a suggested proce- 
dure that may be used to develop an accuracy test for a DTM, 
1. After the completion of the process of reading terrain data 
points from a stereomodel, either the operator or a supervisor 
should read about 25 points scattered throughout that model in 
such a way that the whole area will be sampled, some in drains, 
ridges or sidehills. Special precautions should be taken not 
to read these checkpoints at obvious places where terrain points 
were read, such as on definite breaklines. 
2. Code these checkpoints in such a manner that they will not be 
used to develop the DTM. 
3. Transform all the terrain data and test data to "ground coordi- 
nates" using the same control for both. 
4. The horizontal coordinates of test points are superimposed on 
the DTM and an elevation is interpolated for each point. 
5. Compare the DTM interpolated elevations with the test eleva- 
tions and express the difference in the form of a standard er- 
ror. 
We now have a number which is objective, determined in a systematic 
way and represents the error that one can expect to find in the 
products that will be developed from the DTM.